Keyboard Re-Mapper Utility

By Jörg Langowski, MacTutor Editorial Board

Keyboard configuration and reconfiguration

This time I'd like to address a problem that is rather neglected in Inside Macintosh, but not by MacTutor, and that is the configuration of the Mac keyboard. [As background, you may wish to refer to the August issue for an article by Joel West and David Smith on the Keyboard Sleuth.] The section on international resources mentions the different existing keyboards, their layouts and the keycodes; but it is nowhere said explicitly how the keycode that is sent out when a key is pressed is mapped to an ASCII character and where the tables are kept that are used for that assignment.

There is one program, the Localizer, that is generally used to change the keyboard configuration from one layout to another. This works well for the standard keyboards that come for the different countries in which the Mac is sold. It would be very convenient, though, to be able to change the keyboard mapping according to one's taste. That way one could layout a Dvorak keyboard or make simple changes just to have some different symbols available, without changing the whole keyboard arrangement. As an example, I prefer to have a decimal point instead of a comma on my (German) Mac+ numeric keypad, without changing the rest of the layout so that all the symbols stay where they are. There is no 'canned' configuration in the localizer that provides for this change. So, here's a nice chance to write a small application in Forth that does something useful and at the same time show some instructive details about how the keyboard is configured.

The INIT 0 and INIT 1 resources

Doing so requires some NOSYing around in the routines that translate keycodes into ASCII characters. These routines are built at system initialization time through the INIT resources ID=0 and ID=1. [One can find this out easily by looking at the Localizer, which contains quite a few INIT resources that all look very much the same except for some tables that contain the actual keyboard mapping]. Disassembly reveals how a keycode is translated into an ASCII character by the INIT routines. See Listing 1. The modifier bits (Shift, Option, Caps Lock) are taken as an index into a small offset table, from which the beginning of the actual translation table is obtained. The keycode then serves as an index into that table. Two different translation tables exist for the old Mac and the Mac+ keyboards. The keyboard type is read from the system global KybdType at $21E, which is continually being updated by the background routine that checks for the presence of a keyboard. A hex $B here means we have a Mac+ keyboard plugged in, other numbers so far refer to the old keyboard, and keypad, optionally. [A zero means, there is no keyboard... but in that case, we'll never get to this routine anyway...]

A very simple way to change the translation table to our preferences is to emulate the table lookup process in Forth, starting with the keycode. Mach1 (also the new version Mach2) contains the word ?TERMINAL which returns the character code, keycode and modifier flags if a key has been pressed.

Fig. 1 Our Keyboard Remapper in Mach2

Listing 2 defines a couple of words that will return the absolute address of the translation table entry from the keycode and modifiers. keycode and modifiers return the respective fields. keyaddress, given a modifier / keycode input on the stack, determines which INIT resource to use (INIT 0 for the main keyboard and INIT 1 for the numeric keypad), leaves a handle to that resource on the stack and the table entry's address on top of it. This address can then be used by the remapping routine to put a new entry into the table. The handle will be used to call _ChangedResource to mark the INIT resource as changed so that it is rewritten when _UpdateResource is called at the end.

get.key1.data and get.key2.data are words that are used by keyaddress to determine the beginning of the translation tables for the main and numeric keyboards. They are somewhat kludgy as they make use of the internal structure of the INIT 0 and INIT1 resource codes, reading the table addresses out of the LEA <data>,A0 instructions at the beginning of the code. When a system is released where those instructions occur in different positions, my program will get completely lost. In that case, it would either have to be modified to accommodate the new addresses or the kludge would have to be generalized, using a routine that looks for a LEA <data>,A0 instruction. All this is embarassingly bad programming style, but it was the simplest way that I could think up to find the table addresses.

The INIT resources are read once at the beginning of the main word remap and stored in two variables hand0 and hand1. remap itself is an endless loop which comprises the user dialog for remapping of the keys. When the user hits the go-away box, the loop is exited through a vector which is installed in the task's goaway-hook; the resource file is updated (file id=0 indicates the System file), and control is returned to the Finder.

The program can be compiled into a standalone application through the normal turnkey process built into Mach1; a turnkeyed version can be found on the MacTutor source disk.

While the remapper is being used, the INIT resources already in memory from boot-up will of course stay as they are. Only the copies of INIT 0 and INIT 1 in the present System file are changed and rewritten. This means that your keyboard will stay in its old configuration until you reboot with the disk that contains the changed System file.

Mach2, assembler bug, etc.

The new version of Mach - Mach2 - has been released a long time ago as you read this. As I write this, however, I have just taken a short look at the changes and improvements that have been made. The main change for execution speed is that the DO...LOOP index and limit have been moved from the loop stack into registers (only the innermost loop). This speeds up loop execution tremendously, so that the famous Sieve now executes in just about seven and one half seconds, which is only one second slower than TML Pascal code. Another benchmark I wrote that does 50 screen inversions (explicitly, without calling InvertRect) also executes only about 15% slower than Pascal code.

The disadvantage is that many more registers are now being internally used by the Mach2 system. In fact, Palo Alto Shipping warned me not to use any registers in CODE definitions without saving and restoring them. Therefore, if you try to use the floating point routines or many of the other examples that you could find here with Mach2, you're likely to be disappointed. Inserting MOVEM.L A0-A4/D0-D7,-(A7) and MOVEM.L (A7)+, A0-A4/D0-D7 in strategic places will help tremendously to fight frustration in that case.

The assembler bug which caused faulty coding of some instructions (seeV2#8) has been fixed, and they even promised that the Forth debugger would work Real Soon Now. Meanwhile, TMON is still the best way to debug with Mach2.

Where did the $%&§ output go...

A multitasking system like Mach has its peculiarities, which have to be learned. Somebody from the local developer's group had problems getting Quickdraw calls drawing into a window he had set up in Mach1. When he would turnkey the program, however, everything would work fine. It seemed like under some circumstances, with the Mach1 system window open, the system would use a different grafPort from what he thought.

There is a rather simple explanation to this problem. Mach1 keeps a flashing cursor going in any window in which QUIT, the interpreter main loop, is running, which causes the current port to be reset to that window's grafPort. It is the programmer's responsibility to make sure that the currently active grafPort is the one that one wants to use. Since any task keeps a pointer to its associated window in a user variable, this is easy to achieve by doing a _setPort to this window. (One could argue that the task scheduler should automatically reset the current port to the task's own window. In my opinion, that would not be a good idea, since it would create a lot of overhead on switching tasks. Furthermore, the task actually might want to use another grafPort, this would make the initial resetting unnecessary anyway.)

Listing 1: excerpts from the INIT0 and INIT1 key translation routines
;
; alphanumeric keyboard translator, Apple System File
;
INIT0  BRA     com_1 ; setup jump vector

proc2    BRA.S   lac_2
data1    DC.W    $FFFF,$4B45,$5943,1,$301,0

proc3    LEA     data8,A0    ; Mac+ translate table
 CMPI.B  #11,KybdType ; 11 = Mac+
 BEQ.S   lac_1
  LEA     data9,A0    ; old Mac translate table
lac_1    MOVEQ   #7,D0
 AND     D1,D0
 ADD     D0,D0 ; modifier byte*2
 MOVE    0(A0,D0.W),D0 ; offset for modifiers 
 ADD     D2,D0
 MOVE.B  0(A0,D0.W),D0 ; table entry
 RTS     

lac_2    CMPI    #52,D2 ; main entry point, D2 contains keycode
 BHI.S   lac_3  ; was modifier key pressed?
 BSR     proc3  ; if not, translate through table
 TST.B   KeyMap+6
 BPL.S   lac_4 ; if no control key
 BTST    #0,D1 ; shift pressed?
 BEQ.S   lac_4
 BTST    #2,D1 ; option pressed?
 BNE.S   lac_4
 BCLR    #0,D1 ; clear shift bit
 BSR     proc3  ; and retranslate
 CMPI.B  #48,D0 ; is it
 BCS.S   lac_4 ; a number key?
   CMPI.B  #57,D0
 BHI.S   lac_4
 TST.B   D3
 BMI.S   lac_4
 TST.B   ScrDmpEnb ; deal with FKEYs
 BEQ.S   lac_4
 SUBI.B  #48,D0
 MOVE.B  D0,ScrDmpType
lac_3    MOVEQ   #0,D0
lac_4    .
 .
 . ( code that deals with diacritical marks,
 .  etc. etc. )
 .
 RTS
 .

data8    ; Mac+ key translate table (German keyboard)
 DC.W    $10,$45,$7A,$45,$AF,$E4,$119,$E4
 ; ( offsets for modifier keys)
  DC.W    $6173,$6466,$6867,$7978,$6376,$62,$7177,$6572

  DC.W    $AEC6,$CEB2,$B0D0,$7EB5,$C909,$CAD4,$800

data9    ; Old Mac key translate table (German keyboard)
 DC.W    $10,$45,$7A,$45,$AF,$E4,$119,$E4
 ; ( offsets for modifier keys)
  DC.W  $6173,$6466,$6867,$3C79,$7863,$2D76,$7177, $6572
 .
 . ( DC.W offsets for modifiers etc.)
 .
  
  DC.W    $AEF1,$CE0D,$F7C9,$F5F6,$F809,$3D4,$8CA
                                         
com_1    LEA     proc2,A0
 MOVE.L  A0,Key1Trans  ; $29E
 RTS 
;    
; numeric keypad translator
;
INIT1  BRA     com_1 ; setup jump vector
proc2    LEA     data1,A0    ; translate table
 CMPI.B  #11,KybdType ; 11 = Mac+
 BEQ.S   lac_1
 ADDA.W  #64,A0 ; offset for old Mac
lac_1    LSR     #1,D1
 BCC.S   lac_2
 ADDA.W  #32,A0 ; offset for shift key
lac_2    ANDI.B  #$1F,D2
 MOVE.B  0(A0,D2.W),D0 ; get table entry
 RTS                                             
data1    DC.W    $2C,$1D00,0,$1C1B,$1F00,0,$31E,$2D00
  DC.W    0,$3031,$3233,$3435,$3637,$38,$3900,0
  DC.W    $2E,$2A00,0,$2B1B,$3D00,0,$32F,$2D00
  DC.W    0,$3031,$3233,$3435,$3637,$38,$3900,0
  DC.W    $2C,$1D00,0,$1C1B,$1F00,0,$31E,$2D00
  DC.W    0,$3031,$3233,$3435,$3637,$38,$3900,0
  DC.W    $2C,$2A00,0,$2B1B,$2E00,0,$32F,$2D00
  DC.W    0,$3031,$3233,$3435,$3637,$38,$3900,0
                                         
com_1    LEA     proc2,A0
 MOVE.L  A0,Key2Trans
 RTS                                          
 END     
Listing 2: Reconfiguring the Keyboard routines INIT 0 and INIT 1
( Keyboard reconfigurator, c 1986 J. Langowski / MacTutor )
( For reconfiguration of individual keys without using the Localizer 
)

only forth also assembler also mac

hex
494e4954 constant "init
21e constant keyb.type  ( B: Mac+ )
variable hand0  variable hand1

: get.handles
 "init 0 call getresource hand0 !
 "init 1 call getresource hand1 !
;
: get.key1.data ( | handle start.of.data )
 hand0 @ dup @ 7fffff and 14 + 
 keyb.type c@ B <> if C + then  dup w@ +
;
: get.key2.data ( | handle start.of.data )
 hand1 @ dup @ 7fffff and 6 + dup w@ +
 keyb.type c@ B <> if 40 + then  
;
: keyin
 begin 2d emit 8 emit ?terminal ?dup until
;
: keycode ( extracts keycode out of keyinfo )
 100 / ff and ;

: modifiers ( extracts modifiers from keyinfo )
 1000000 / ff and ;

: keyaddress { | mods keyc -- handle offset }
 dup  modifiers -> mods  keycode -> keyc
 keyc 41 < if get.key1.data 
 dup mods E and + w@ + keyc + 
   else get.key2.data 
 mods 2 and
 if ( shift key down) 20 + then
 keyc 1f and +
   then
;
: keytest
 begin keyin keyaddress c@ emit key drop drop again 
;
: get&show ( | handle offset )
 keyin dup key emit ."  - Keycode = " 
 dup modifiers swap keycode . 
 ." , Modifiers = " binary 
 <# # # # # #> type decimal 
 keyaddress dup c@ 
 ." , actual mapping - " emit cr
;
: keynum cr 
 begin get&show drop drop again
;
decimal

: hello 
 cls
 ." Macintosh Keyboard Remapper" cr
 ." © 1986 J.Langowski / MacTutor" cr cr
 ." Closing the window will update the resource file." cr cr
;
: yesno begin key dup emit 
 case 
 89 of 1 1 endof  121 of 1 1 endof
 78 of 0 1 endof  110 of 0 1 endof
 cr ." y or n - " 0 swap
 endcase
   until
;
: on.goaway 0 call updateresfile bye ;
 
: remap
 hello 
 get.handles
 begin
 ." Type the key you'd like to change - " cr
 get&show ( | handle offset )
 ." Do you want to remap this key (y/n) - " yesno cr
 if 
 ." Enter the character to map this key to - "
 key dup emit cr 
 swap c! 
     ( store in mapping table of INIT resource)
 call changedresource
 else 2drop
 then cr
 again
;
 
164 user goaway-hook

NEW.WINDOW remapper
" Keyboard Configuration Editor" remapper TITLE
50 20 316 496 remapper BOUNDS
ROUNDED VISIBLE CLOSEBOX NOGROWBOX 
 remapper ITEMS
600 5000 terminal mapper

: go.map activate remap ;

: start
 remapper add
 remapper mapper build
 ['] on.goaway goaway-hook task-> mapper !
 remapper dup call selectwindow call setport
 mapper go.map
;